fix(grok): surface context window usage in the composer

[ahmed-besic]

What Changed

Grok already reports session context fill and model window size over ACP, but T3 never turned that into the composer context meter that Codex already shows.

This maps:

• ACP sessionUpdate: "usage_update" (used / size) when present
• Grok's current path: session/update_meta.totalTokens while streaming, plus availableModels[]._meta.totalContextTokens for the window size
• Prompt response _meta.totalTokens / _meta.usage as a reliable end-of-turn snapshot

Those become thread.token-usage.updated → context-window.updated, which the existing ContextWindowMeter already renders. No new UI components.

Cursor gets a no-op case for the shared TokenUsageUpdated parse event so the ACP union stays exhaustive; wiring makeAcpTokenUsageUpdatedEvent there is the same one-liner once verified against a real Cursor ACP install.

Why

Codex threads show a context ring (used / max / %). Grok threads showed nothing, even though grok agent stdio already streams fill (_meta.totalTokens) and advertises per-model windows (totalContextTokens, e.g. 500k / 200k / 128k). Without the adapter mapping, the meter never got a snapshot.

Live probe against grok 0.2.118: Grok does not emit standard ACP usage_update yet, so this PR supports that path for when it lands and uses the _meta sources that work today.

UI Changes

Uses the existing context window meter in the composer.

• Before: Grok threads have no context size / fill indicator.
• After: After the first usage-bearing update (stream meta or prompt result), Grok shows the same ring + used/max tooltip as Codex.

Validation

• vp test run src/provider/acp/AcpRuntimeModel.test.ts src/provider/acp/AcpCoreRuntimeEvents.test.ts src/provider/acp/GrokAcpSupport.test.ts — 23 passed
• vp test run src/provider/Layers/GrokAdapter.test.ts — 20 passed
• Live ACP probe (grok agent stdio): confirmed _meta.totalTokens on session/update, totalContextTokens on models, and prompt _meta.usage; no usage_update from this CLI version yet
• Meter math from live numbers: e.g. 19,267 / 500,000 ≈ 3.9% used

Checklist

• This PR is small and focused
• I explained what changed and why
• I included before/after screenshots for any UI changes (happy to add after a UI smoke against a Grok thread)
• A video is not applicable (no animation/timing change)

Made by Grok Build via the Grok Build harness while working on T3 Code.

---

Note

Medium Risk
Touches Grok ACP session event handling and token-usage snapshots that feed the composer meter. Parsing is additive and conservative (positive used tokens, no guessing among multiple windows), but incorrect mapping would show a wrong fill % to users.

Overview
Grok threads now feed the existing composer context meter. The adapter emits thread.token-usage.updated from ACP usage_update (when present), streaming session/update_meta.totalTokens, and prompt-response _meta as an end-of-turn snapshot, enriched with per-model totalContextTokens and compactsAutomatically.

Session context keeps a modelId→window map and last-known max so fill % still works between turns and after model lookup misses. Usage events are allowed after activeTurnId is cleared so the meter can update between turns; interrupted turns are skipped.

Cursor only adds a no-op TokenUsageUpdated case so the shared ACP event union stays exhaustive. UI change is a grok → Grok display-name mapping.

^{Reviewed by Cursor Bugbot for commit 8a491a6. Bugbot is set up for automated code reviews on this repo. Configure here.}

Note

Surface context window usage in the Grok provider composer

• Adds thread.token-usage.updated event emission to the Grok adapter, fired both during streaming (via ACP session/update notifications) and at the end of each prompt turn from prompt response metadata.
• Introduces helpers in GrokAcpSupport.ts to extract context window sizes from model metadata, build a modelId→window map, and enrich usage snapshots with maxTokens and compactsAutomatically.
• Extends AcpRuntimeModel.ts to parse usage_update ACP notifications and _meta fields into TokenUsageUpdated events.
• Adds makeAcpTokenUsageUpdatedEvent factory in AcpCoreRuntimeEvents.ts for consistent construction of token-usage runtime events across ACP providers.
• Maps the grok provider string to 'Grok' in contextWindow.ts for correct UI display.

^{Macroscope summarized 8a491a6.}

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 943c338f-0f2a-40ce-9163-8bef728f6237

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[macroscopeapp]

Approvability

Verdict: Needs human review

This PR introduces new user-facing functionality (context window usage meter in the composer) with ~650 lines of new code, new event types, and new state management. Despite the 'fix' label, this is a feature addition that warrants human review.

^{You can customize Macroscope's approvability policy. Learn more.}

[ahmed-besic]

Merged current main. Context-window tests, Grok adapter tests, and server/web typechecks are passing.

[theman6660]

Live check against current Grok ACP: tokenUsageFromGrokPromptMeta should not feed the composer meter.

On a greeting turn:

• session/update_meta.totalTokens = 1615 (context occupancy)
• Prompt response usage.totalTokens = 20554 (billed tokens)

If the prompt-response path runs after the notification, the ring jumps to the billed number. Grok still does not send usage_update; occupancy lives on session/update _meta.totalTokens. Window size came from session signals.json (contextWindowTokens: 500000), not from the prompt result.

The rest of the mapping (accept usage_update if it ever lands, plus _meta.totalTokens while streaming) matches what I saw. The prompt-meta snapshot is the part that would show the wrong fill %.

[ahmed-besic]

What Changed

Grok already reports session context fill and model window size over ACP, but T3 never turned that into the composer context meter that Codex already shows.

This maps:

• ACP sessionUpdate: "usage_update" (used / size) when present
• Grok's current path: session/update_meta.totalTokens while streaming, plus availableModels[]._meta.totalContextTokens for the window size
• Prompt response _meta.totalTokens / _meta.usage as a reliable end-of-turn snapshot

Those become thread.token-usage.updated → context-window.updated, which the existing ContextWindowMeter already renders. No new UI components.

Cursor gets a no-op case for the shared TokenUsageUpdated parse event so the ACP union stays exhaustive; wiring makeAcpTokenUsageUpdatedEvent there is the same one-liner once verified against a real Cursor ACP install.

Why

Codex threads show a context ring (used / max / %). Grok threads showed nothing, even though grok agent stdio already streams fill (_meta.totalTokens) and advertises per-model windows (totalContextTokens, e.g. 500k / 200k / 128k). Without the adapter mapping, the meter never got a snapshot.

Live probe against grok 0.2.118: Grok does not emit standard ACP usage_update yet, so this PR supports that path for when it lands and uses the _meta sources that work today.

UI Changes

Uses the existing context window meter in the composer.

• Before: Grok threads have no context size / fill indicator.
• After: After the first usage-bearing update (stream meta or prompt result), Grok shows the same ring + used/max tooltip as Codex.

Validation

• vp test run src/provider/acp/AcpRuntimeModel.test.ts src/provider/acp/AcpCoreRuntimeEvents.test.ts src/provider/acp/GrokAcpSupport.test.ts — 23 passed
• vp test run src/provider/Layers/GrokAdapter.test.ts — 20 passed
• Live ACP probe (grok agent stdio): confirmed _meta.totalTokens on session/update, totalContextTokens on models, and prompt _meta.usage; no usage_update from this CLI version yet
• Meter math from live numbers: e.g. 19,267 / 500,000 ≈ 3.9% used

Checklist

• This PR is small and focused
• I explained what changed and why
• I included before/after screenshots for any UI changes (happy to add after a UI smoke against a Grok thread)
• A video is not applicable (no animation/timing change)

Made by Grok Build via the Grok Build harness while working on T3 Code.

---

Note

Medium Risk
Touches Grok ACP session event handling and token-usage snapshots that feed the composer meter. Parsing is additive and conservative (positive used tokens, no guessing among multiple windows), but incorrect mapping would show a wrong fill % to users.

Overview
Grok threads now feed the existing composer context meter. The adapter emits thread.token-usage.updated from ACP usage_update (when present), streaming session/update_meta.totalTokens, and prompt-response _meta as an end-of-turn snapshot, enriched with per-model totalContextTokens and compactsAutomatically.

Session context keeps a modelId→window map and last-known max so fill % still works between turns and after model lookup misses. Usage events are allowed after activeTurnId is cleared so the meter can update between turns; interrupted turns are skipped.

Cursor only adds a no-op TokenUsageUpdated case so the shared ACP event union stays exhaustive. UI change is a grok → Grok display-name mapping.

^{Reviewed by Cursor Bugbot for commit 8a491a6. Bugbot is set up for automated code reviews on this repo. Configure here.}

Note

Surface context window usage in the Grok provider composer

• Adds thread.token-usage.updated event emission to the Grok adapter, fired both during streaming (via ACP session/update notifications) and at the end of each prompt turn from prompt response metadata.
• Introduces helpers in GrokAcpSupport.ts to extract context window sizes from model metadata, build a modelId→window map, and enrich usage snapshots with maxTokens and compactsAutomatically.
• Extends AcpRuntimeModel.ts to parse usage_update ACP notifications and _meta fields into TokenUsageUpdated events.
• Adds makeAcpTokenUsageUpdatedEvent factory in AcpCoreRuntimeEvents.ts for consistent construction of token-usage runtime events across ACP providers.
• Maps the grok provider string to 'Grok' in contextWindow.ts for correct UI display.

^{Macroscope summarized 8a491a6.}

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 943c338f-0f2a-40ce-9163-8bef728f6237

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[macroscopeapp]

Approvability

Verdict: Needs human review

This PR introduces new user-facing functionality (context window usage meter in the composer) with ~650 lines of new code, new event types, and new state management. Despite the 'fix' label, this is a feature addition that warrants human review.

^{You can customize Macroscope's approvability policy. Learn more.}

[ahmed-besic]

Merged current main. Context-window tests, Grok adapter tests, and server/web typechecks are passing.

[theman6660]

Live check against current Grok ACP: tokenUsageFromGrokPromptMeta should not feed the composer meter.

On a greeting turn:

• session/update_meta.totalTokens = 1615 (context occupancy)
• Prompt response usage.totalTokens = 20554 (billed tokens)

If the prompt-response path runs after the notification, the ring jumps to the billed number. Grok still does not send usage_update; occupancy lives on session/update _meta.totalTokens. Window size came from session signals.json (contextWindowTokens: 500000), not from the prompt result.

The rest of the mapping (accept usage_update if it ever lands, plus _meta.totalTokens while streaming) matches what I saw. The prompt-meta snapshot is the part that would show the wrong fill %.

[ahmed-besic]

What Changed

Grok already reports session context fill and model window size over ACP, but T3 never turned that into the composer context meter that Codex already shows.

This maps:

• ACP sessionUpdate: "usage_update" (used / size) when present
• Grok's current path: session/update_meta.totalTokens while streaming, plus availableModels[]._meta.totalContextTokens for the window size
• Prompt response _meta.totalTokens / _meta.usage as a reliable end-of-turn snapshot

Those become thread.token-usage.updated → context-window.updated, which the existing ContextWindowMeter already renders. No new UI components.

Cursor gets a no-op case for the shared TokenUsageUpdated parse event so the ACP union stays exhaustive; wiring makeAcpTokenUsageUpdatedEvent there is the same one-liner once verified against a real Cursor ACP install.

Why

Codex threads show a context ring (used / max / %). Grok threads showed nothing, even though grok agent stdio already streams fill (_meta.totalTokens) and advertises per-model windows (totalContextTokens, e.g. 500k / 200k / 128k). Without the adapter mapping, the meter never got a snapshot.

Live probe against grok 0.2.118: Grok does not emit standard ACP usage_update yet, so this PR supports that path for when it lands and uses the _meta sources that work today.

UI Changes

Uses the existing context window meter in the composer.

• Before: Grok threads have no context size / fill indicator.
• After: After the first usage-bearing update (stream meta or prompt result), Grok shows the same ring + used/max tooltip as Codex.

Validation

• vp test run src/provider/acp/AcpRuntimeModel.test.ts src/provider/acp/AcpCoreRuntimeEvents.test.ts src/provider/acp/GrokAcpSupport.test.ts — 23 passed
• vp test run src/provider/Layers/GrokAdapter.test.ts — 20 passed
• Live ACP probe (grok agent stdio): confirmed _meta.totalTokens on session/update, totalContextTokens on models, and prompt _meta.usage; no usage_update from this CLI version yet
• Meter math from live numbers: e.g. 19,267 / 500,000 ≈ 3.9% used

Checklist

• This PR is small and focused
• I explained what changed and why
• I included before/after screenshots for any UI changes (happy to add after a UI smoke against a Grok thread)
• A video is not applicable (no animation/timing change)

Made by Grok Build via the Grok Build harness while working on T3 Code.

---

Note

Medium Risk
Touches Grok ACP session event handling and token-usage snapshots that feed the composer meter. Parsing is additive and conservative (positive used tokens, no guessing among multiple windows), but incorrect mapping would show a wrong fill % to users.

Overview
Grok threads now feed the existing composer context meter. The adapter emits thread.token-usage.updated from ACP usage_update (when present), streaming session/update_meta.totalTokens, and prompt-response _meta as an end-of-turn snapshot, enriched with per-model totalContextTokens and compactsAutomatically.

Session context keeps a modelId→window map and last-known max so fill % still works between turns and after model lookup misses. Usage events are allowed after activeTurnId is cleared so the meter can update between turns; interrupted turns are skipped.

Cursor only adds a no-op TokenUsageUpdated case so the shared ACP event union stays exhaustive. UI change is a grok → Grok display-name mapping.

^{Reviewed by Cursor Bugbot for commit 8a491a6. Bugbot is set up for automated code reviews on this repo. Configure here.}

Note

Surface context window usage in the Grok provider composer

• Adds thread.token-usage.updated event emission to the Grok adapter, fired both during streaming (via ACP session/update notifications) and at the end of each prompt turn from prompt response metadata.
• Introduces helpers in GrokAcpSupport.ts to extract context window sizes from model metadata, build a modelId→window map, and enrich usage snapshots with maxTokens and compactsAutomatically.
• Extends AcpRuntimeModel.ts to parse usage_update ACP notifications and _meta fields into TokenUsageUpdated events.
• Adds makeAcpTokenUsageUpdatedEvent factory in AcpCoreRuntimeEvents.ts for consistent construction of token-usage runtime events across ACP providers.
• Maps the grok provider string to 'Grok' in contextWindow.ts for correct UI display.

^{Macroscope summarized 8a491a6.}

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 943c338f-0f2a-40ce-9163-8bef728f6237

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[macroscopeapp]

Approvability

Verdict: Needs human review

This PR introduces new user-facing functionality (context window usage meter in the composer) with ~650 lines of new code, new event types, and new state management. Despite the 'fix' label, this is a feature addition that warrants human review.

^{You can customize Macroscope's approvability policy. Learn more.}

[ahmed-besic]

Merged current main. Context-window tests, Grok adapter tests, and server/web typechecks are passing.

[theman6660]

Live check against current Grok ACP: tokenUsageFromGrokPromptMeta should not feed the composer meter.

On a greeting turn:

• session/update_meta.totalTokens = 1615 (context occupancy)
• Prompt response usage.totalTokens = 20554 (billed tokens)

If the prompt-response path runs after the notification, the ring jumps to the billed number. Grok still does not send usage_update; occupancy lives on session/update _meta.totalTokens. Window size came from session signals.json (contextWindowTokens: 500000), not from the prompt result.

The rest of the mapping (accept usage_update if it ever lands, plus _meta.totalTokens while streaming) matches what I saw. The prompt-meta snapshot is the part that would show the wrong fill %.

[ahmed-besic]

What Changed

Grok already reports session context fill and model window size over ACP, but T3 never turned that into the composer context meter that Codex already shows.

This maps:

• ACP sessionUpdate: "usage_update" (used / size) when present
• Grok's current path: session/update_meta.totalTokens while streaming, plus availableModels[]._meta.totalContextTokens for the window size
• Prompt response _meta.totalTokens / _meta.usage as a reliable end-of-turn snapshot

Those become thread.token-usage.updated → context-window.updated, which the existing ContextWindowMeter already renders. No new UI components.

Cursor gets a no-op case for the shared TokenUsageUpdated parse event so the ACP union stays exhaustive; wiring makeAcpTokenUsageUpdatedEvent there is the same one-liner once verified against a real Cursor ACP install.

Why

Codex threads show a context ring (used / max / %). Grok threads showed nothing, even though grok agent stdio already streams fill (_meta.totalTokens) and advertises per-model windows (totalContextTokens, e.g. 500k / 200k / 128k). Without the adapter mapping, the meter never got a snapshot.

Live probe against grok 0.2.118: Grok does not emit standard ACP usage_update yet, so this PR supports that path for when it lands and uses the _meta sources that work today.

UI Changes

Uses the existing context window meter in the composer.

• Before: Grok threads have no context size / fill indicator.
• After: After the first usage-bearing update (stream meta or prompt result), Grok shows the same ring + used/max tooltip as Codex.

Validation

• vp test run src/provider/acp/AcpRuntimeModel.test.ts src/provider/acp/AcpCoreRuntimeEvents.test.ts src/provider/acp/GrokAcpSupport.test.ts — 23 passed
• vp test run src/provider/Layers/GrokAdapter.test.ts — 20 passed
• Live ACP probe (grok agent stdio): confirmed _meta.totalTokens on session/update, totalContextTokens on models, and prompt _meta.usage; no usage_update from this CLI version yet
• Meter math from live numbers: e.g. 19,267 / 500,000 ≈ 3.9% used

Checklist

• This PR is small and focused
• I explained what changed and why
• I included before/after screenshots for any UI changes (happy to add after a UI smoke against a Grok thread)
• A video is not applicable (no animation/timing change)

Made by Grok Build via the Grok Build harness while working on T3 Code.

---

Note

Medium Risk
Touches Grok ACP session event handling and token-usage snapshots that feed the composer meter. Parsing is additive and conservative (positive used tokens, no guessing among multiple windows), but incorrect mapping would show a wrong fill % to users.

Overview
Grok threads now feed the existing composer context meter. The adapter emits thread.token-usage.updated from ACP usage_update (when present), streaming session/update_meta.totalTokens, and prompt-response _meta as an end-of-turn snapshot, enriched with per-model totalContextTokens and compactsAutomatically.

Session context keeps a modelId→window map and last-known max so fill % still works between turns and after model lookup misses. Usage events are allowed after activeTurnId is cleared so the meter can update between turns; interrupted turns are skipped.

Cursor only adds a no-op TokenUsageUpdated case so the shared ACP event union stays exhaustive. UI change is a grok → Grok display-name mapping.

^{Reviewed by Cursor Bugbot for commit 8a491a6. Bugbot is set up for automated code reviews on this repo. Configure here.}

Note

Surface context window usage in the Grok provider composer

• Adds thread.token-usage.updated event emission to the Grok adapter, fired both during streaming (via ACP session/update notifications) and at the end of each prompt turn from prompt response metadata.
• Introduces helpers in GrokAcpSupport.ts to extract context window sizes from model metadata, build a modelId→window map, and enrich usage snapshots with maxTokens and compactsAutomatically.
• Extends AcpRuntimeModel.ts to parse usage_update ACP notifications and _meta fields into TokenUsageUpdated events.
• Adds makeAcpTokenUsageUpdatedEvent factory in AcpCoreRuntimeEvents.ts for consistent construction of token-usage runtime events across ACP providers.
• Maps the grok provider string to 'Grok' in contextWindow.ts for correct UI display.

^{Macroscope summarized 8a491a6.}

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 943c338f-0f2a-40ce-9163-8bef728f6237

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[macroscopeapp]

Approvability

Verdict: Needs human review

This PR introduces new user-facing functionality (context window usage meter in the composer) with ~650 lines of new code, new event types, and new state management. Despite the 'fix' label, this is a feature addition that warrants human review.

^{You can customize Macroscope's approvability policy. Learn more.}

[ahmed-besic]

Merged current main. Context-window tests, Grok adapter tests, and server/web typechecks are passing.

[theman6660]

Live check against current Grok ACP: tokenUsageFromGrokPromptMeta should not feed the composer meter.

On a greeting turn:

• session/update_meta.totalTokens = 1615 (context occupancy)
• Prompt response usage.totalTokens = 20554 (billed tokens)

If the prompt-response path runs after the notification, the ring jumps to the billed number. Grok still does not send usage_update; occupancy lives on session/update _meta.totalTokens. Window size came from session signals.json (contextWindowTokens: 500000), not from the prompt result.

The rest of the mapping (accept usage_update if it ever lands, plus _meta.totalTokens while streaming) matches what I saw. The prompt-meta snapshot is the part that would show the wrong fill %.

[ahmed-besic]

What Changed

Grok already reports session context fill and model window size over ACP, but T3 never turned that into the composer context meter that Codex already shows.

This maps:

• ACP sessionUpdate: "usage_update" (used / size) when present
• Grok's current path: session/update_meta.totalTokens while streaming, plus availableModels[]._meta.totalContextTokens for the window size
• Prompt response _meta.totalTokens / _meta.usage as a reliable end-of-turn snapshot

Those become thread.token-usage.updated → context-window.updated, which the existing ContextWindowMeter already renders. No new UI components.

Cursor gets a no-op case for the shared TokenUsageUpdated parse event so the ACP union stays exhaustive; wiring makeAcpTokenUsageUpdatedEvent there is the same one-liner once verified against a real Cursor ACP install.

Why

Codex threads show a context ring (used / max / %). Grok threads showed nothing, even though grok agent stdio already streams fill (_meta.totalTokens) and advertises per-model windows (totalContextTokens, e.g. 500k / 200k / 128k). Without the adapter mapping, the meter never got a snapshot.

Live probe against grok 0.2.118: Grok does not emit standard ACP usage_update yet, so this PR supports that path for when it lands and uses the _meta sources that work today.

UI Changes

Uses the existing context window meter in the composer.

• Before: Grok threads have no context size / fill indicator.
• After: After the first usage-bearing update (stream meta or prompt result), Grok shows the same ring + used/max tooltip as Codex.

Validation

• vp test run src/provider/acp/AcpRuntimeModel.test.ts src/provider/acp/AcpCoreRuntimeEvents.test.ts src/provider/acp/GrokAcpSupport.test.ts — 23 passed
• vp test run src/provider/Layers/GrokAdapter.test.ts — 20 passed
• Live ACP probe (grok agent stdio): confirmed _meta.totalTokens on session/update, totalContextTokens on models, and prompt _meta.usage; no usage_update from this CLI version yet
• Meter math from live numbers: e.g. 19,267 / 500,000 ≈ 3.9% used

Checklist

• This PR is small and focused
• I explained what changed and why
• I included before/after screenshots for any UI changes (happy to add after a UI smoke against a Grok thread)
• A video is not applicable (no animation/timing change)

Made by Grok Build via the Grok Build harness while working on T3 Code.

---

Note

Medium Risk
Touches Grok ACP session event handling and token-usage snapshots that feed the composer meter. Parsing is additive and conservative (positive used tokens, no guessing among multiple windows), but incorrect mapping would show a wrong fill % to users.

Overview
Grok threads now feed the existing composer context meter. The adapter emits thread.token-usage.updated from ACP usage_update (when present), streaming session/update_meta.totalTokens, and prompt-response _meta as an end-of-turn snapshot, enriched with per-model totalContextTokens and compactsAutomatically.

Session context keeps a modelId→window map and last-known max so fill % still works between turns and after model lookup misses. Usage events are allowed after activeTurnId is cleared so the meter can update between turns; interrupted turns are skipped.

Cursor only adds a no-op TokenUsageUpdated case so the shared ACP event union stays exhaustive. UI change is a grok → Grok display-name mapping.

^{Reviewed by Cursor Bugbot for commit 8a491a6. Bugbot is set up for automated code reviews on this repo. Configure here.}

Note

Surface context window usage in the Grok provider composer

• Adds thread.token-usage.updated event emission to the Grok adapter, fired both during streaming (via ACP session/update notifications) and at the end of each prompt turn from prompt response metadata.
• Introduces helpers in GrokAcpSupport.ts to extract context window sizes from model metadata, build a modelId→window map, and enrich usage snapshots with maxTokens and compactsAutomatically.
• Extends AcpRuntimeModel.ts to parse usage_update ACP notifications and _meta fields into TokenUsageUpdated events.
• Adds makeAcpTokenUsageUpdatedEvent factory in AcpCoreRuntimeEvents.ts for consistent construction of token-usage runtime events across ACP providers.
• Maps the grok provider string to 'Grok' in contextWindow.ts for correct UI display.

^{Macroscope summarized 8a491a6.}

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 943c338f-0f2a-40ce-9163-8bef728f6237

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[macroscopeapp]

Approvability

Verdict: Needs human review

This PR introduces new user-facing functionality (context window usage meter in the composer) with ~650 lines of new code, new event types, and new state management. Despite the 'fix' label, this is a feature addition that warrants human review.

^{You can customize Macroscope's approvability policy. Learn more.}

[ahmed-besic]

Merged current main. Context-window tests, Grok adapter tests, and server/web typechecks are passing.

[theman6660]

Live check against current Grok ACP: tokenUsageFromGrokPromptMeta should not feed the composer meter.

On a greeting turn:

• session/update_meta.totalTokens = 1615 (context occupancy)
• Prompt response usage.totalTokens = 20554 (billed tokens)

If the prompt-response path runs after the notification, the ring jumps to the billed number. Grok still does not send usage_update; occupancy lives on session/update _meta.totalTokens. Window size came from session signals.json (contextWindowTokens: 500000), not from the prompt result.

The rest of the mapping (accept usage_update if it ever lands, plus _meta.totalTokens while streaming) matches what I saw. The prompt-meta snapshot is the part that would show the wrong fill %.

[ahmed-besic]

What Changed

Grok already reports session context fill and model window size over ACP, but T3 never turned that into the composer context meter that Codex already shows.

This maps:

• ACP sessionUpdate: "usage_update" (used / size) when present
• Grok's current path: session/update_meta.totalTokens while streaming, plus availableModels[]._meta.totalContextTokens for the window size
• Prompt response _meta.totalTokens / _meta.usage as a reliable end-of-turn snapshot

Those become thread.token-usage.updated → context-window.updated, which the existing ContextWindowMeter already renders. No new UI components.

Cursor gets a no-op case for the shared TokenUsageUpdated parse event so the ACP union stays exhaustive; wiring makeAcpTokenUsageUpdatedEvent there is the same one-liner once verified against a real Cursor ACP install.

Why

Codex threads show a context ring (used / max / %). Grok threads showed nothing, even though grok agent stdio already streams fill (_meta.totalTokens) and advertises per-model windows (totalContextTokens, e.g. 500k / 200k / 128k). Without the adapter mapping, the meter never got a snapshot.

Live probe against grok 0.2.118: Grok does not emit standard ACP usage_update yet, so this PR supports that path for when it lands and uses the _meta sources that work today.

UI Changes

Uses the existing context window meter in the composer.

• Before: Grok threads have no context size / fill indicator.
• After: After the first usage-bearing update (stream meta or prompt result), Grok shows the same ring + used/max tooltip as Codex.

Validation

• vp test run src/provider/acp/AcpRuntimeModel.test.ts src/provider/acp/AcpCoreRuntimeEvents.test.ts src/provider/acp/GrokAcpSupport.test.ts — 23 passed
• vp test run src/provider/Layers/GrokAdapter.test.ts — 20 passed
• Live ACP probe (grok agent stdio): confirmed _meta.totalTokens on session/update, totalContextTokens on models, and prompt _meta.usage; no usage_update from this CLI version yet
• Meter math from live numbers: e.g. 19,267 / 500,000 ≈ 3.9% used

Checklist

• This PR is small and focused
• I explained what changed and why
• I included before/after screenshots for any UI changes (happy to add after a UI smoke against a Grok thread)
• A video is not applicable (no animation/timing change)

Made by Grok Build via the Grok Build harness while working on T3 Code.

---

Note

Medium Risk
Touches Grok ACP session event handling and token-usage snapshots that feed the composer meter. Parsing is additive and conservative (positive used tokens, no guessing among multiple windows), but incorrect mapping would show a wrong fill % to users.

Overview
Grok threads now feed the existing composer context meter. The adapter emits thread.token-usage.updated from ACP usage_update (when present), streaming session/update_meta.totalTokens, and prompt-response _meta as an end-of-turn snapshot, enriched with per-model totalContextTokens and compactsAutomatically.

Session context keeps a modelId→window map and last-known max so fill % still works between turns and after model lookup misses. Usage events are allowed after activeTurnId is cleared so the meter can update between turns; interrupted turns are skipped.

Cursor only adds a no-op TokenUsageUpdated case so the shared ACP event union stays exhaustive. UI change is a grok → Grok display-name mapping.

^{Reviewed by Cursor Bugbot for commit 8a491a6. Bugbot is set up for automated code reviews on this repo. Configure here.}

Note

Surface context window usage in the Grok provider composer

• Adds thread.token-usage.updated event emission to the Grok adapter, fired both during streaming (via ACP session/update notifications) and at the end of each prompt turn from prompt response metadata.
• Introduces helpers in GrokAcpSupport.ts to extract context window sizes from model metadata, build a modelId→window map, and enrich usage snapshots with maxTokens and compactsAutomatically.
• Extends AcpRuntimeModel.ts to parse usage_update ACP notifications and _meta fields into TokenUsageUpdated events.
• Adds makeAcpTokenUsageUpdatedEvent factory in AcpCoreRuntimeEvents.ts for consistent construction of token-usage runtime events across ACP providers.
• Maps the grok provider string to 'Grok' in contextWindow.ts for correct UI display.

^{Macroscope summarized 8a491a6.}

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 943c338f-0f2a-40ce-9163-8bef728f6237

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[macroscopeapp]

Approvability

Verdict: Needs human review

This PR introduces new user-facing functionality (context window usage meter in the composer) with ~650 lines of new code, new event types, and new state management. Despite the 'fix' label, this is a feature addition that warrants human review.

^{You can customize Macroscope's approvability policy. Learn more.}

[ahmed-besic]

Merged current main. Context-window tests, Grok adapter tests, and server/web typechecks are passing.

[theman6660]

Live check against current Grok ACP: tokenUsageFromGrokPromptMeta should not feed the composer meter.

On a greeting turn:

• session/update_meta.totalTokens = 1615 (context occupancy)
• Prompt response usage.totalTokens = 20554 (billed tokens)

If the prompt-response path runs after the notification, the ring jumps to the billed number. Grok still does not send usage_update; occupancy lives on session/update _meta.totalTokens. Window size came from session signals.json (contextWindowTokens: 500000), not from the prompt result.

The rest of the mapping (accept usage_update if it ever lands, plus _meta.totalTokens while streaming) matches what I saw. The prompt-meta snapshot is the part that would show the wrong fill %.

[ahmed-besic]

What Changed

Grok already reports session context fill and model window size over ACP, but T3 never turned that into the composer context meter that Codex already shows.

This maps:

• ACP sessionUpdate: "usage_update" (used / size) when present
• Grok's current path: session/update_meta.totalTokens while streaming, plus availableModels[]._meta.totalContextTokens for the window size
• Prompt response _meta.totalTokens / _meta.usage as a reliable end-of-turn snapshot

Those become thread.token-usage.updated → context-window.updated, which the existing ContextWindowMeter already renders. No new UI components.

Cursor gets a no-op case for the shared TokenUsageUpdated parse event so the ACP union stays exhaustive; wiring makeAcpTokenUsageUpdatedEvent there is the same one-liner once verified against a real Cursor ACP install.

Why

Codex threads show a context ring (used / max / %). Grok threads showed nothing, even though grok agent stdio already streams fill (_meta.totalTokens) and advertises per-model windows (totalContextTokens, e.g. 500k / 200k / 128k). Without the adapter mapping, the meter never got a snapshot.

Live probe against grok 0.2.118: Grok does not emit standard ACP usage_update yet, so this PR supports that path for when it lands and uses the _meta sources that work today.

UI Changes

Uses the existing context window meter in the composer.

• Before: Grok threads have no context size / fill indicator.
• After: After the first usage-bearing update (stream meta or prompt result), Grok shows the same ring + used/max tooltip as Codex.

Validation

• vp test run src/provider/acp/AcpRuntimeModel.test.ts src/provider/acp/AcpCoreRuntimeEvents.test.ts src/provider/acp/GrokAcpSupport.test.ts — 23 passed
• vp test run src/provider/Layers/GrokAdapter.test.ts — 20 passed
• Live ACP probe (grok agent stdio): confirmed _meta.totalTokens on session/update, totalContextTokens on models, and prompt _meta.usage; no usage_update from this CLI version yet
• Meter math from live numbers: e.g. 19,267 / 500,000 ≈ 3.9% used

Checklist

• This PR is small and focused
• I explained what changed and why
• I included before/after screenshots for any UI changes (happy to add after a UI smoke against a Grok thread)
• A video is not applicable (no animation/timing change)

Made by Grok Build via the Grok Build harness while working on T3 Code.

---

Note

Medium Risk
Touches Grok ACP session event handling and token-usage snapshots that feed the composer meter. Parsing is additive and conservative (positive used tokens, no guessing among multiple windows), but incorrect mapping would show a wrong fill % to users.

Overview
Grok threads now feed the existing composer context meter. The adapter emits thread.token-usage.updated from ACP usage_update (when present), streaming session/update_meta.totalTokens, and prompt-response _meta as an end-of-turn snapshot, enriched with per-model totalContextTokens and compactsAutomatically.

Session context keeps a modelId→window map and last-known max so fill % still works between turns and after model lookup misses. Usage events are allowed after activeTurnId is cleared so the meter can update between turns; interrupted turns are skipped.

Cursor only adds a no-op TokenUsageUpdated case so the shared ACP event union stays exhaustive. UI change is a grok → Grok display-name mapping.

^{Reviewed by Cursor Bugbot for commit 8a491a6. Bugbot is set up for automated code reviews on this repo. Configure here.}

Note

Surface context window usage in the Grok provider composer

• Adds thread.token-usage.updated event emission to the Grok adapter, fired both during streaming (via ACP session/update notifications) and at the end of each prompt turn from prompt response metadata.
• Introduces helpers in GrokAcpSupport.ts to extract context window sizes from model metadata, build a modelId→window map, and enrich usage snapshots with maxTokens and compactsAutomatically.
• Extends AcpRuntimeModel.ts to parse usage_update ACP notifications and _meta fields into TokenUsageUpdated events.
• Adds makeAcpTokenUsageUpdatedEvent factory in AcpCoreRuntimeEvents.ts for consistent construction of token-usage runtime events across ACP providers.
• Maps the grok provider string to 'Grok' in contextWindow.ts for correct UI display.

^{Macroscope summarized 8a491a6.}

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 943c338f-0f2a-40ce-9163-8bef728f6237

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[macroscopeapp]

Approvability

Verdict: Needs human review

This PR introduces new user-facing functionality (context window usage meter in the composer) with ~650 lines of new code, new event types, and new state management. Despite the 'fix' label, this is a feature addition that warrants human review.

^{You can customize Macroscope's approvability policy. Learn more.}

[ahmed-besic]

Merged current main. Context-window tests, Grok adapter tests, and server/web typechecks are passing.

[theman6660]

Live check against current Grok ACP: tokenUsageFromGrokPromptMeta should not feed the composer meter.

On a greeting turn:

• session/update_meta.totalTokens = 1615 (context occupancy)
• Prompt response usage.totalTokens = 20554 (billed tokens)

If the prompt-response path runs after the notification, the ring jumps to the billed number. Grok still does not send usage_update; occupancy lives on session/update _meta.totalTokens. Window size came from session signals.json (contextWindowTokens: 500000), not from the prompt result.

The rest of the mapping (accept usage_update if it ever lands, plus _meta.totalTokens while streaming) matches what I saw. The prompt-meta snapshot is the part that would show the wrong fill %.

[ahmed-besic]

What Changed

Grok already reports session context fill and model window size over ACP, but T3 never turned that into the composer context meter that Codex already shows.

This maps:

• ACP sessionUpdate: "usage_update" (used / size) when present
• Grok's current path: session/update_meta.totalTokens while streaming, plus availableModels[]._meta.totalContextTokens for the window size
• Prompt response _meta.totalTokens / _meta.usage as a reliable end-of-turn snapshot

Those become thread.token-usage.updated → context-window.updated, which the existing ContextWindowMeter already renders. No new UI components.

Cursor gets a no-op case for the shared TokenUsageUpdated parse event so the ACP union stays exhaustive; wiring makeAcpTokenUsageUpdatedEvent there is the same one-liner once verified against a real Cursor ACP install.

Why

Codex threads show a context ring (used / max / %). Grok threads showed nothing, even though grok agent stdio already streams fill (_meta.totalTokens) and advertises per-model windows (totalContextTokens, e.g. 500k / 200k / 128k). Without the adapter mapping, the meter never got a snapshot.

Live probe against grok 0.2.118: Grok does not emit standard ACP usage_update yet, so this PR supports that path for when it lands and uses the _meta sources that work today.

UI Changes

Uses the existing context window meter in the composer.

• Before: Grok threads have no context size / fill indicator.
• After: After the first usage-bearing update (stream meta or prompt result), Grok shows the same ring + used/max tooltip as Codex.

Validation

• vp test run src/provider/acp/AcpRuntimeModel.test.ts src/provider/acp/AcpCoreRuntimeEvents.test.ts src/provider/acp/GrokAcpSupport.test.ts — 23 passed
• vp test run src/provider/Layers/GrokAdapter.test.ts — 20 passed
• Live ACP probe (grok agent stdio): confirmed _meta.totalTokens on session/update, totalContextTokens on models, and prompt _meta.usage; no usage_update from this CLI version yet
• Meter math from live numbers: e.g. 19,267 / 500,000 ≈ 3.9% used

Checklist

• This PR is small and focused
• I explained what changed and why
• I included before/after screenshots for any UI changes (happy to add after a UI smoke against a Grok thread)
• A video is not applicable (no animation/timing change)

Made by Grok Build via the Grok Build harness while working on T3 Code.

---

Note

Medium Risk
Touches Grok ACP session event handling and token-usage snapshots that feed the composer meter. Parsing is additive and conservative (positive used tokens, no guessing among multiple windows), but incorrect mapping would show a wrong fill % to users.

Overview
Grok threads now feed the existing composer context meter. The adapter emits thread.token-usage.updated from ACP usage_update (when present), streaming session/update_meta.totalTokens, and prompt-response _meta as an end-of-turn snapshot, enriched with per-model totalContextTokens and compactsAutomatically.

Session context keeps a modelId→window map and last-known max so fill % still works between turns and after model lookup misses. Usage events are allowed after activeTurnId is cleared so the meter can update between turns; interrupted turns are skipped.

Cursor only adds a no-op TokenUsageUpdated case so the shared ACP event union stays exhaustive. UI change is a grok → Grok display-name mapping.

^{Reviewed by Cursor Bugbot for commit 8a491a6. Bugbot is set up for automated code reviews on this repo. Configure here.}

Note

Surface context window usage in the Grok provider composer

• Adds thread.token-usage.updated event emission to the Grok adapter, fired both during streaming (via ACP session/update notifications) and at the end of each prompt turn from prompt response metadata.
• Introduces helpers in GrokAcpSupport.ts to extract context window sizes from model metadata, build a modelId→window map, and enrich usage snapshots with maxTokens and compactsAutomatically.
• Extends AcpRuntimeModel.ts to parse usage_update ACP notifications and _meta fields into TokenUsageUpdated events.
• Adds makeAcpTokenUsageUpdatedEvent factory in AcpCoreRuntimeEvents.ts for consistent construction of token-usage runtime events across ACP providers.
• Maps the grok provider string to 'Grok' in contextWindow.ts for correct UI display.

^{Macroscope summarized 8a491a6.}

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 943c338f-0f2a-40ce-9163-8bef728f6237

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[macroscopeapp]

Approvability

Verdict: Needs human review

This PR introduces new user-facing functionality (context window usage meter in the composer) with ~650 lines of new code, new event types, and new state management. Despite the 'fix' label, this is a feature addition that warrants human review.

^{You can customize Macroscope's approvability policy. Learn more.}

[ahmed-besic]

Merged current main. Context-window tests, Grok adapter tests, and server/web typechecks are passing.

[theman6660]

Live check against current Grok ACP: tokenUsageFromGrokPromptMeta should not feed the composer meter.

On a greeting turn:

• session/update_meta.totalTokens = 1615 (context occupancy)
• Prompt response usage.totalTokens = 20554 (billed tokens)

If the prompt-response path runs after the notification, the ring jumps to the billed number. Grok still does not send usage_update; occupancy lives on session/update _meta.totalTokens. Window size came from session signals.json (contextWindowTokens: 500000), not from the prompt result.

The rest of the mapping (accept usage_update if it ever lands, plus _meta.totalTokens while streaming) matches what I saw. The prompt-meta snapshot is the part that would show the wrong fill %.